[RFC] 9533: sixth iteration

[Paolo Carlini]

Pétur Runólfsson wrote:

>This looks great, but if this is meant for 3.3 this:
>
>+	  if (__cvt.encoding() >= 0)
>+	    __ret += _M_file.showmanyc_helper(__sync) / __cvt.max_length();
>
>should be changed to:
>
>+	  if (__cvt.always_noconv())
>+	    __ret += _M_file.showmanyc_helper(__sync);
>
>since codecvt<wchar_t, char, mbstate_t>::encoding() and max_length() in
>3.3 return garbage (they are not perfect in 3.4 either :-).
>
Actually, the patch was tested with 3.4: I'm reluctant to apply 
something first to 3.3, then to trunk.

Thanks for the info, anyway, which will be much useful in the near future.

What do you mean exactly by "garbage" (*)? Could you provide some more 
info or, even better, a patch, enabling us to use safely the much more 
correct code on top?

Thanks,
Paolo.

(*) Your colorful language often reminds me that of H.J. Lu ;-)